/Instagram-post-analysis

Analysis of Instagram Post Engagement through Topic Modeling

Primary LanguageJupyter NotebookMIT LicenseMIT

Analysis of Instagram Post Engagement through Topic Modeling

1. Introduction

This report presents an analysis of Instagram post engagement through topic modeling of labels derived from images using Google Vision. The analysis is based on a sample of 296 posts containing a total of 1522 images. The goal is to identify which topics are associated with higher engagement, defined as the number of comments. This information can guide content strategy to increase engagement on the platform.

2. Methodology

Labels were first obtained for each image using Google Vision API, the dataset was then subjected to topic modeling to extract prevalent themes from the labels obtained via Google Vision on each image. Latent Dirichlet Allocation (LDA) was employed to categorize the captions into eight distinct topics. Each image was assigned a distribution across these topics. For efficiency, not all images were analyzed. Instead, a representative sample was chosen to expedite the computation and ensure manageability.

3. Topic Descriptions

Below are the topics extracted from the analysis, with key words representing each topic:

  • Formal Events and Fashion: Smile, Sleeve, Flash photography, Formal wear.
  • Outdoor Leisure and Vision: Eyewear, Sunglasses, People in nature, Vision care.
  • Casual Wear and Happy Moments: Gesture, Happy, Smile, T-shirt.
  • Musical Events and Performance: Event, Font, Black-and-white, Microphone.
  • Sports and Athletic Wear: Sports uniform, Jersey, Gesture, Player.
  • Digital Content and Branding: Font, Event, Electric blue, Brand.
  • Nature and Outdoors: Plant, Happy, Sky, Tree.
  • Food and Cuisine: Food, Ingredient, Recipe, Cuisine.

4. Engagement Analysis

The following table shows the average topic weights for posts with high and low engagement. In other words, the probabibilties of posts, whether high engagement or low engagement, to belong to a topic. High engagement posts are those in the top 25% of the number of comments garnered, while low engagement posts fall in the bottom 25%.

Topic High Engagement Avg Weights Low Engagement Avg Weights
Formal Events and Fashion 0.152796 0.130055
Outdoor Leisure and Vision 0.135640 0.123097
Casual Wear and Happy Moments 0.112544 0.153359
Musical Events and Performance 0.097433 0.099123
Sports and Athletic Wear 0.182938 0.127017
Digital Content and Branding 0.092226 0.162322
Nature and Outdoors 0.129024 0.096579
Food and Cuisine 0.097399 0.108449

5. Recommendations

Based on the analysis, the following strategies are recommended to increase engagement:

  • Increase the frequency of posts related to Sports and Athletic Wear and Formal Events and Fashions, as these topics are strongly associated with high engagement.
  • Consider diversifying content to include more topics related to Nature and Outdoors and Outdoor Leisure and Vision which also show potential for high engagement.
  • Review the content strategy for Casual Wear and Happy Moments, Digital Content and Branding, and Food and Cuisine as these topics show higher weights in low engagement posts.

6. Conclusion

This analysis provides a clear indication of which topics are most likely to engage an Instagram audience. By strategically focusing on these topics, it is possible to enhance overall engagement on the platform.